fix: allow custom image when calling deploy or create_model with various frameworks (#1347)

Author: mattmcclean

src/sagemaker/chainer/estimator.py

@@ -215,7 +215,7 @@ def create_model(
             py_version=self.py_version,
             framework_version=self.framework_version,
             model_server_workers=model_server_workers,
-            image=self.image_name,
+            image=kwargs["image"] if "image" in kwargs else self.image_name,
             sagemaker_session=self.sagemaker_session,
             vpc_config=self.get_vpc_config(vpc_config_override),
             dependencies=(dependencies or self.dependencies),

src/sagemaker/mxnet/estimator.py

@@ -210,7 +210,7 @@ def create_model(
             code_location=self.code_location,
             py_version=self.py_version,
             framework_version=self.framework_version,
-            image=(image_name or self.image_name),
+            image=kwargs["image"] if "image" in kwargs else (image_name or self.image_name),
             model_server_workers=model_server_workers,
             sagemaker_session=self.sagemaker_session,
             vpc_config=self.get_vpc_config(vpc_config_override),

src/sagemaker/pytorch/estimator.py

@@ -175,7 +175,7 @@ def create_model(
             code_location=self.code_location,
             py_version=self.py_version,
             framework_version=self.framework_version,
-            image=self.image_name,
+            image=kwargs["image"] if "image" in kwargs else self.image_name,
             model_server_workers=model_server_workers,
             sagemaker_session=self.sagemaker_session,
             vpc_config=self.get_vpc_config(vpc_config_override),

src/sagemaker/rl/estimator.py

@@ -218,7 +218,7 @@ def create_model(
         base_args = dict(
             model_data=self.model_data,
             role=role or self.role,
-            image=self.image_name,
+            image=kwargs["image"] if "image" in kwargs else self.image_name,
             name=self._current_job_name,
             container_log_level=self.container_log_level,
             sagemaker_session=self.sagemaker_session,

src/sagemaker/sklearn/estimator.py

@@ -167,6 +167,13 @@ def create_model(
             logger.debug("removing unused entry_point argument: %s", str(kwargs["entry_point"]))
             kwargs = {k: v for k, v in kwargs.items() if k != "entry_point"}
 
+        # remove image kwarg
+        if "image" in kwargs:
+            image = kwargs["image"]
+            kwargs = {k: v for k, v in kwargs.items() if k != "image"}
+        else:
+            image = None
+
         return SKLearnModel(
             self.model_data,
             role,
@@ -179,7 +186,7 @@ def create_model(
             py_version=self.py_version,
             framework_version=self.framework_version,
             model_server_workers=model_server_workers,
-            image=self.image_name,
+            image=image or self.image_name,
             sagemaker_session=self.sagemaker_session,
             vpc_config=self.get_vpc_config(vpc_config_override),
             enable_network_isolation=self.enable_network_isolation(),

src/sagemaker/tensorflow/estimator.py

@@ -601,10 +601,17 @@ def _create_tfs_model(
         **kwargs
     ):
         """Placeholder docstring"""
+        # remove image kwarg
+        if "image" in kwargs:
+            image = kwargs["image"]
+            kwargs = {k: v for k, v in kwargs.items() if k != "image"}
+        else:
+            image = None
+
         return Model(
             model_data=self.model_data,
             role=role,
-            image=self.image_name,
+            image=(image or self.image_name),
             name=self._current_job_name,
             container_log_level=self.container_log_level,
             framework_version=utils.get_short_version(self.framework_version),
@@ -628,14 +635,21 @@ def _create_default_model(
         **kwargs
     ):
         """Placeholder docstring"""
+        # remove image kwarg
+        if "image" in kwargs:
+            image = kwargs["image"]
+            kwargs = {k: v for k, v in kwargs.items() if k != "image"}
+        else:
+            image = None
+
         return TensorFlowModel(
             self.model_data,
             role,
             entry_point or self.entry_point,
             source_dir=source_dir or self._model_source_dir(),
             enable_cloudwatch_metrics=self.enable_cloudwatch_metrics,
             env={"SAGEMAKER_REQUIREMENTS": self.requirements_file},
-            image=self.image_name,
+            image=(image or self.image_name),
             name=self._current_job_name,
             container_log_level=self.container_log_level,
             code_location=self.code_location,

tests/unit/test_chainer.py

@@ -642,3 +642,11 @@ def test_model_empty_framework_version(warning, sagemaker_session):
     )
     assert model.framework_version == defaults.CHAINER_VERSION
     warning.assert_called_with(defaults.CHAINER_VERSION, defaults.LATEST_VERSION)
+
+
+def test_custom_image_estimator_deploy(sagemaker_session):
+    custom_image = "mycustomimage:latest"
+    chainer = _chainer_estimator(sagemaker_session)
+    chainer.fit(inputs="s3://mybucket/train", job_name="new_name")
+    model = chainer.create_model(image=custom_image)
+    assert model.image == custom_image

tests/unit/test_mxnet.py

@@ -834,3 +834,17 @@ def test_mx_enable_sm_metrics_if_fw_ver_is_at_least_1_6(sagemaker_session):
             framework_version=fw_version,
         )
         assert mx.enable_sagemaker_metrics
+
+
+def test_custom_image_estimator_deploy(sagemaker_session):
+    custom_image = "mycustomimage:latest"
+    mx = MXNet(
+        entry_point=SCRIPT_PATH,
+        role=ROLE,
+        sagemaker_session=sagemaker_session,
+        train_instance_count=INSTANCE_COUNT,
+        train_instance_type=INSTANCE_TYPE,
+    )
+    mx.fit(inputs="s3://mybucket/train", job_name="new_name")
+    model = mx.create_model(image=custom_image)
+    assert model.image == custom_image

tests/unit/test_pytorch.py

@@ -602,3 +602,11 @@ def test_pt_enable_sm_metrics_if_fw_ver_is_at_least_1_15(sagemaker_session):
     for fw_version in ["1.3", "1.4", "2.0", "2.1"]:
         pytorch = _pytorch_estimator(sagemaker_session, framework_version=fw_version)
         assert pytorch.enable_sagemaker_metrics
+
+
+def test_custom_image_estimator_deploy(sagemaker_session):
+    custom_image = "mycustomimage:latest"
+    pytorch = _pytorch_estimator(sagemaker_session)
+    pytorch.fit(inputs="s3://mybucket/train", job_name="new_name")
+    model = pytorch.create_model(image=custom_image)
+    assert model.image == custom_image

tests/unit/test_rl.py

@@ -618,3 +618,11 @@ def test_wrong_type_parameters(sagemaker_session):
             train_instance_type=INSTANCE_TYPE,
         )
     assert "combination is not supported." in str(e.value)
+
+
+def test_custom_image_estimator_deploy(sagemaker_session):
+    custom_image = "mycustomimage:latest"
+    rl = _rl_estimator(sagemaker_session)
+    rl.fit(inputs="s3://mybucket/train", job_name="new_name")
+    model = rl.create_model(image=custom_image)
+    assert model.image == custom_image

tests/unit/test_sklearn.py

@@ -570,3 +570,11 @@ def test_model_py2_warning(warning, sagemaker_session):
     )
     assert model.py_version == "py2"
     warning.assert_called_with(model.__framework_name__, defaults.LATEST_PY2_VERSION)
+
+
+def test_custom_image_estimator_deploy(sagemaker_session):
+    custom_image = "mycustomimage:latest"
+    sklearn = _sklearn_estimator(sagemaker_session)
+    sklearn.fit(inputs="s3://mybucket/train", job_name="new_name")
+    model = sklearn.create_model(image=custom_image)
+    assert model.image == custom_image

tests/unit/test_tf_estimator.py

@@ -1300,3 +1300,11 @@ def test_tf_enable_sm_metrics_if_fw_ver_is_at_least_1_15(sagemaker_session):
     for fw_version in ["1.15", "1.16", "2.0", "2.1"]:
         tf = _build_tf(sagemaker_session, framework_version=fw_version)
         assert tf.enable_sagemaker_metrics
+
+
+def test_custom_image_estimator_deploy(sagemaker_session):
+    custom_image = "mycustomimage:latest"
+    tf = _build_tf(sagemaker_session)
+    tf.fit(inputs="s3://mybucket/train", job_name="new_name")
+    model = tf.create_model(image=custom_image)
+    assert model.image == custom_image